Research, vision and conversion

In addition to conventional text sources, Alcilla can prepare websites, images, and Office documents for further work. These functions require additional services or suitable AI models and must therefore be enabled selectively.

Web research

When web search through SearXNG is configured, the research area becomes available.

Research can proceed step by step:

1. analyze the question,
2. plan a search strategy,
3. find web sources,
4. read relevant pages,
5. summarize intermediate findings,
6. investigate open questions in greater depth,
7. create a structured report.

Web research does not replace expert evaluation of sources. The availability, currency, and quality of external websites are outside Alcilla’s control.

Which connections are made?

Research accesses the configured SearXNG instance and then the websites found. This function is optional. Without a configured web search, the research area does not appear in Alcilla.

Vision: understand content in images

The Vision area can analyze images using a multimodal model. Possible applications include:

• capturing text from screenshots or photographs,
• describing visible content,
• analyzing document images,
• providing structured summaries of information from diagrams or user interfaces.

A vision-capable model is required. The specific model is selected in Alcilla and can differ from the model used for ordinary text chat.

Result quality depends heavily on image resolution, language, subject matter, and the selected model. For important documents, the extracted result should always be compared with the original image.

Convert documents

The Convert area prepares Office and PDF documents for the knowledge index. It uses the Docling service.

The process leaves your original files unchanged:

• Additional Markdown files are created.
• Source and destination folders are kept separate.
• New or modified documents can be processed again.
• Progress and logs are visible.

Typical use cases include:

• making existing Word manuals available to knowledge search,
• searching PDF procedures together with emails,
• gradually importing project archives into a knowledge base,
• making the content of email attachments accessible,
• performing supplementary analysis of scanned or image-based content.

Setup assistance

Your internal team does not need prior knowledge of SearXNG, Docling, or multimodal models. We can assist with setup when required.